Abstract PS4-18: Real-world PD-L1 test utilization and analytical concordance of the PD-L1 IHC 28-8 and 22C3 assays in patients with breast cancer

Abstract

Abstract Background: Programmed death-1/programmed death ligand 1 (PD-[L]1) inhibitors are approved for use in a range of cancers. PD-L1 expression in the tumor microenvironment, assessed with an FDA-approved PD-L1 immunohistochemistry (IHC) diagnostic assay such as the Dako PD-L1 IHC 28-8 and 22C3 pharmDx or Ventana PD-L1 SP142 and SP263 assays, is associated with improved PD-(L)1 inhibitor treatment outcomes in some tumor types, including breast cancer (BC). In March 2019, the FDA approved atezolizumab + nab-paclitaxel for the treatment of patients with advanced triple-negative BC and immune cell (IC) PD-L1 expression ≥ 1% using the SP142 assay. Here, we investigated test utilization, test turnaround time (TAT), PD-L1 expression prevalence by assay and biopsy location, and analytical concordance between assays in real-world BC samples. Design: The study included samples from patients with BC that were tested for PD-L1 expression between Oct 2015 and Sep 2019 at NeoGenomics Laboratories, a US national reference laboratory. Patient characteristics from Symphony Healthcare Solutions were matched to PD-L1 test results using unique identifiers. Test volume and TAT were assessed for the 28-8, 22C3, SP142, and SP263 assays. PD-L1 expression was determined by trained pathologists using the 28-8, 22C3, or SP142 assays. Results for the 28-8 assay for the entire study period and for the 22C3 assay until Dec 2018 were reported as the percentage of tumor cells (% TC) with PD-L1 expression. From Jan 2019 onwards, 22C3 assay results were reported as a combined positive score (CPS). All SP142 assay results were reported as the percentage of ICs (% IC) with PD-L1 expression. Analytical concordance between assays was assessed in patients with matched samples (biopsies from the same site and collected on the same date). BioStat Solutions performed statistical analyses. Results: 2955 PD-L1 tests were performed on samples from 2508 patients with BC. The volume of PD-L1 tests on BC samples increased > 100-fold over the study period. Mean TAT was < 5 days for all 4 assays pooled. Table 1 shows PD-L1 expression prevalence in patients with a 28-8, 22C3, or SP142 test result. Median PD-L1 expression did not differ between primary tumors and metastatic sites. In matched samples, overall percentage agreement (OPA) between the 28-8 (TC ≥ 1%) and 22C3 (CPS ≥ 1) assays was 94%, and OPA between the 22C3 (CPS ≥ 1) and SP142 (IC ≥ 1%) assays was 64% (Table 2). Analytical concordance between the 28-8 and 22C3 assays for % TC scoring in matched samples from 27 patients was high (Kendall’s tau = 0.997 [95% CI, 0.883-1.000]). Conclusion: Mean PD-L1 test TAT for BC samples remained < 5 days across all tests despite a large increase in test volume over the study period. Prevalence of PD-L1 expression ≥ 1(%) was higher with the CPS and % IC algorithms than the % TC algorithm, although differences could be due to multiple confounding factors. Despite a small sample size, analytical concordance between the 28-8 and 22C3 assays in matched samples was high. These data provide real-world context for the PD-L1 testing landscape in BC. Table 1. Prevalence of PD-L1 expression in patients with BCPD-L1 expressiona28-8 and 22C3b22C3cSP142% TC, n (%)CPS, n (%)% IC, n (%)(N = 608)(N = 609)(N = 1080)<1(%)390 (64)253 (42)367 (34)≥1(%)218 (36)356 (58)713 (66)All patients had a single test result or ≥ 2 identical results. aThe CPS algorithm is reported on a scale of 0-100, not as a percentage; bSamples tested with the 22C3 assay between Q4 2015 and Q4 2018 were scored using the % TC algorithm; cSamples tested with the 22C3 assay between Q1 2019 and Q4 2019 were scored using the CPS algorithm. Table 2. Agreement between assays on matched samples from patients with BCAgreement between 28-8 (TC ≥ 1%) and 22C3 (CPS ≥ 1) (N = 18)a28-8 as reference22C3 as referenceOPA (n/N)94 (17/18)PPA (n/N)100 (6/6)86 (6/7)NPA (n/N)92 (11/12)100 (11/11)Agreement between 22C3 (CPS ≥ 1) and SP142 (IC ≥ 1%) (N = 33)b22C3 as referenceSP142 as referenceOPA (n/N)64 (21/33)PPA (n/N)86 (12/14)55 (12/22)NPA (n/N)47 (9/19)82 (9/11)aData are presented for 28-8 and 22C3 tests on matched samples with 22C3 tests performed between Q1 2019 and Q4 2019; bData are presented for 22C3 and SP142 tests on matched samples with 22C3 tests performed between Q1 2019 and Q4 2019. N, total number of samples; n, number of samples with the same results with the 2 tests; NPA, negative percentage agreement; PPA, positive percentage agreement. Citation Format: Shreya Mitra, Emily A. Prince, James Pratt, James Novotny, Jr, Vladislav Chizhevsky, Josette William Ragheb, David Huron. Real-world PD-L1 test utilization and analytical concordance of the PD-L1 IHC 28-8 and 22C3 assays in patients with breast cancer [abstract]. In: Proceedings of the 2020 San Antonio Breast Cancer Virtual Symposium; 2020 Dec 8-11; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2021;81(4 Suppl):Abstract nr PS4-18.